Blog Post: The MGA Model Was Built for the AI Boom

By Maggie Shaw, Head of Business Development & Partnerships, LEEO Insurance

I came into insurance from fintech and SaaS, where the tool you shipped in January was usually ready for upgrades by June. Insurance doesn't work like that, and for the most part that's a feature, not a bug. It's old, it's necessary, and it pays people's claims on the worst day of their year. You don't actually want the industry covering your commercial fleet to be the kind that moves fast and breaks things.

But that carefulness has a cost, and AI is making the bill come due faster than usual. The tools are changing every few months. The regulations are being rewritten while we're still reading the last draft. Whatever model you picked at the start of the year is already a little dated. A big carrier can't reasonably rip out its underwriting stack every quarter to keep up. Honestly, it shouldn't.

That gap is exactly why MGAs exist, and it's why I think the MGA model is quietly the best-built structure in P&C for this particular moment.

Quick refresher if "MGA" isn't your daily vocabulary: a managing general agent underwrites on behalf of a carrier or reinsurer that supplies the capacity. We take on the underwriting, the distribution, and the technology. They take on the risk and protect their balance sheet. It's an arrangement that has stopped being a footnote — U.S. MGA premium hit roughly $114 billion in 2024, up 16% year over year, and it's overwhelmingly a property-and-casualty story. The structure is asset-light and fast, without the institutional weight that makes incumbents slow.

Here's why that fits the AI boom so well.

Capacity partners get to outsource the risk of going first. Someone has to be the one testing the bleeding-edge tooling, and most carriers and reinsurers would prefer it not be them. The MGA becomes the sandbox. We try the new model, the new data feed, the new workflow. If it works, our capacity partners get exposure to the upside of moving fast. If it flops, they didn't build a department around it. They get to evaluate the future without having to commit to it sight unseen.

You can test technology across lines without betting the company. Because we're light, we can run a new tool on one slice of the book, watch how it actually performs, and either roll it out or quietly kill it. My bread and butter, commercial auto, is a great place to start because it's one of the most data-rich lines in all of P&C. Telematics and sensors mean you're not guessing; you can prove out AI-driven underwriting on real signal and then port what works to adjacent lines. And when the rules or the tools shift halfway through the year, adapting looks more like a config change than a reorg.

It lowers the cost of trying lines that look unattractive but aren't. Some lines look ugly on paper — thin loss history, messy to price — so carriers skip them, even when there's real upside for whoever figures them out first. The MGA structure makes being wrong cheap. You can take the swing on an unproven line without standing up an entire carrier apparatus to do it, and AI is quietly making more of those swings worth taking.

Put those together and you get something the industry doesn't usually have: a way to move at the speed of the technology without asking a centuries-old institution to behave like a startup. P&C normally absorbs this kind of change over decades. The MGA model does it in quarters.

That's the part I find genuinely interesting. The standard story about insurance and technology is that the industry is a laggard and always will be. But the MGA model is a structural answer to that: a way for careful, necessary, balance-sheet-protecting institutions to keep a foot in whatever's coming next. It lets the slow part stay slow where it should, and the fast part move where it can.